RE: Configuring GCC build: doc patch

["Dave Korn" <dk at artimi dot com>]

> -----Original Message-----
> From: Hugh Sasse Staff Elec Eng [mailto:hgs@dmu.ac.uk] 
> Sent: 19 July 2004 14:30

> On Mon, 19 Jul 2004, Dave Korn wrote:
> 
>    [my suggestion "Do not use a relative path for configure" elided]
> >
> >  Works fine for me.  I think it's only certain hosts that have this
> > problem, isn't it?
> 
> My case was Solaris 9 sparc 

  Ah yes, that was the one: came up only a couple of weeks ago.

http://gcc.gnu.org/ml/gcc/2004-07/msg00025.html

> I don't know why it should fail in such a case. 

  Some kind of discrepancy between Solaris shell semantics and other
systems, at a guess, but I don't know.

> I can't really speak about other systems.

  As far as I know, it's thoroughly reliable on most of the other mainstream
gcc platforms; Solaris is the only context I've heard anyone have problems
with it, although there's one post in the thread I referenced above where
someone thought they might once have seen the same on freebsd5.2.  But for
the rest, it seems to be perfectly acceptable to invoke configure with
either a relative or absolute path.  

  So the doc patch you suggested belongs more properly in the
target-specific build/install instructions at
http://gcc.gnu.org/install/specific.html, and in fact, if you check at
http://gcc.gnu.org/install/specific.html#_002a_002d_002a_002dsolaris2_002a,
you will see that someone's already documented this problem:

"The Solaris 2 /bin/sh will often fail to configure libstdc++-v3, boehm-gc
or libjava. We therefore recommend to use the following sequence of commands
to bootstrap and install GCC: 

        % CONFIG_SHELL=/bin/ksh
        % export CONFIG_SHELL

and then proceed as described in the build instructions, where we strongly
recommend using GNU make and specifying an absolute path to invoke
srcdir/configure. "

although the wording there makes it sound like it's only a problem for
configuring some of the subdirs; maybe the best approach would be just to
modify the test above to make it clear that the same problem can manifest as
other problems, including at "make install" time.



    cheers, 
      DaveK
-- 
Can't think of a witty .sigline today....